Reversible Control of Native GluN2B-Containing NMDA Receptors with Visible Light

NMDA receptors (NMDARs) are glutamate-gated ion channels playing a central role in synaptic transmission and plasticity. NMDAR dysregulation is linked to various neuropsychiatric disorders. This is particularly true for GluN2B-containing NMDARs (GluN2B-NMDARs), which have major pro-cognitive, but also pro-excitotoxic roles, although their exact involvement in these processes remains debated. Traditional GluN2B-selective antagonists suffer from slow and irreversible effects, limiting their use in native tissues. We therefore developed OptoNAM-3, a photoswitchable negative allosteric modulator selective for GluN2B-NMDARs. OptoNAM-3 provided light-induced reversible inhibition of GluN2B-NMDAR activity with precise temporal control both in vitro and in vivo on the behavior of freely moving Xenopus tadpoles. When bound to GluN2B-NMDARs, OptoNAM-3 displayed remarkable red-shifting of its photoswitching properties allowing the use of blue light instead of UV light to turn-off its activity, which we attributed to geometric constraints imposed by the binding site onto the azobenzene moiety of the ligand. This study therefore highlights the importance of the binding site in shaping the photochemical properties of azobenzene-based photoswitches. In addition, by enabling selective, fast, and reversible photocontrol of native GluN2B-NMDARs with in vivo compatible photochemical properties (visible light), OptoNAM-3 should be a useful tool for the investigation of the GluN2B-NMDAR physiology in native tissues.

To characterize the photochemical properties of OptoNAM-1, -2 and -4, we acquired UV-visible spectra of these compounds diluted in physiological buffer (see Methods) in the dark (trans state) and after illumination with various wavelengths.The absorption spectra of OptoNAM-1, -2 and 4 in the dark were characteristic of azobenzenes in their trans configuration 1 (black curves in Figure S1B,F,J).Application of UV light at a wavelength close to the main absorption peak of the trans form (365 nm) gave a completely different spectrum (violet curves in Figure S1B,F,J), characteristic of azobenzenes in their cis configuration 1 .
To determine the appropriate wavelengths to convert the cis OptoNAMs back to trans, we illuminated the 365 nm PSS OptoNAM-1, -2 and -4 solutions with wavelengths of 440 and 490 nm.440 nm irradiation provided the strongest cis-to-trans conversion for OptoNAM-1 and -4 (OptoNAM-1, ~75% trans measured at λ trans = 320 nm; OptoNAM-4, 100% trans measured at λ trans = 334 nm), while 490 nm was most efficient to convert cis-OptoNAM-2 back to trans (~52% measured at λ trans = 371 nm) (Figure S1A,B,E,F,I,J).The cis isomers of all compounds displayed strong thermal stability since the absorption spectrum of the 365 nm PSS kept in the dark did not change over 1 h at room temperature (Figure S1C,G,K).
The activities of the dark and 365 nm PSS of OptoNAM-1, -2 and -4 on GluN1/GluN2B NMDARs were assessed by electrophysiology on Xenopus oocytes as described in the main text for OptoNAM-3.
OptoNAM-1 and -2 behaved as GluN1/GluN2B NAMs with better apparent affinity in the dark than in the UV condition.However, OptoNAM-1 and 2 in trans configuration displayed a drastic loss of potency compared to their parent compounds, [2][3][4][5][6] with a >1000-fold shift in potency (Figure S1D,H and TableS1).We showed that replacement of the amino-methyl bond of the parent compounds by an azo bond to obtain OptoNAM-1 and -2 induced a loss of protonation of the aminopyridium moiety at physiological pH, which likely disrupts binding of the compounds in the ifenprodil binding site (FigureS2).OptoNAM-4, on the other hand, did not display any photodependence of activity (Figure S1 and Table S1).
1 µs with no constraints (Traj01-03 in Figure S7A).They all started from the same structure, but were assigned different initial random velocities during the equilibration procedure (see Methods).We followed the evolution of two dihedral angles during these one that describes the orientation of the NH from aniline (Figure 7A), and one that describes the orientation of one of the azo group relative to the central azobenzene phenyl (C-C-N=N dihedral angle; orientation of azo; Figure 7A).During these simulations, the C-N=N-C angles from the azo group stayed at 180° i.e.OptoNAM-3 did not convert from trans to cis.We observed however that both the aniline and C-C-N=N angles could interconvert between 0° and 180° (Figure S7A).Thus, the ligand is flexible in the active site with different possible conformations, and in particular different conformations of the terminal phenyl-azo moiety.We furthermore observed that the protein was also quite flexible.Indeed, the distance between GluN1 and GluN2B NTD lower lobes significantly increased during the simulation (35 Å between the Cα of GluN1-K179 and GluN2B-K185 at the end of the simulations vs 25 Å in the 5EWJ crystal structure).This can easily be explained by the fact that we are simulating a dimer of isolated GluN1 and GluN2B NTDs, whose relative orientation is usually constrained by the rest of the protein in the full-length receptor. 8We thus decided to perform new simulations where the protein heavy atoms were restrained close to the crystallographic position by a soft harmonic potential of 100 kJ/mol/nm 2 .One simulation started from the same initial conformation of OptoNAM-3 as before (C-C-N=N angle of 141°, called Rot-1; Figure 7A), whereas in the other we manually rotated the C-C-N=N dihedral angle by 180° (going from 141° to -39°, Rot-2 rotamer; Figure 7A) to reflect the OptoNAM-3 conformational mobility observed in the first round of simulations.
To study the behavior of free trans-OptoNAM-3, we performed a 4 µs-long MD simulation of the ligand alone in water.We observed transitions of both the aniline and the C-C-N=N angle all along the trajectory (Figure S7C).Compared to the simulations performed in the protein, we observed much more transitions between each conformation in solution, which means that the free energy barrier to go from one conformation to the other is much smaller in solution than when bound to the protein.(M,N) OptoNAM-3 trans-to-cis and cis-to-trans isomerization for different illumination wavelengths.(M) OptoNAM-3 UV-visible absorption spectra in physiological aqueous buffer (Ringer pH 7.3) in the dark (black curve) and PSS obtained after illumination with wavelengths ranging from 350 to 635 nm of the dark PSS.These spectra were used to create panel D and E from Figure 4. (N) OptoNAM-3 UV-visible absorption spectra in the dark (black curve), after 365 nm illumination (violet curve), and PSS obtained after illumination with wavelengths ranging from 350 to 635 nm of the 365 nm PSS.These spectra were used to create panels I and J from Figure 5.A leftward shift and a decrease of the absorbance peak corresponding to the aminopyridine carrying the charge was observed upon OptoNAM deprotonation.This effect is most obvious for OptoNAM-2 and characteristic of the deprotonated spectra of 2,6-diaminopyridine bases. 9Based on this analysis, at physiological pH, OptoNAM-1 and -2, either in cis or trans, are unprotonated, while OptoNAM-3 is protonated.
(G) Relationship between the activity and the measured pKa of compounds from the same chemical series as parent compound 1 (values from ref. 3) (black dots).Parent compound 1 is highlighted as a thick green dot.Parent compound 2 (orange), as well as trans-OptoNAM-1 (red) and -2 (grey) were added to the plot according to their published or measured activity (Supplementary Table 1), and predicted pKa (pKa was predicted by Marvin, Chemaxon https://www.chemaxon.com)."Neutral" (dark green) and "Positive" (light green) indicate whether the compounds of a given pKa are neutral (dark green) or positively charged (light green) at physiological pH (pH = 7.3).Note the tight correlation between pKa and activity, suggesting that the decreased pKa of OptoNAM-1 and -2 induced by azologization of the parent compounds, resulting in a loss of protonation at physiological pH, is responsible for the large decrease of activity of these compounds.Percentage of neuronal survival in cultured cortical neurons exposed either to control (0.01% DMSO and 10 µM Glycine), NMDA (100 µM NMDA + 10 µM Glycine), NMDA + ifenprodil or NMDA + OptoNAM-3 (100 µM NMDA + 10 μM Glycine + 5 µM of inhibitor), in the dark (grey bar) or after 2 min UV (365 nm) illumination (violet bar).In the presence of trans-OptoNAM-3 (dark) or ifenprodil, cell survival increased to 50% and 60%, respectively.When 365 nm illumination followed the addition of OptoNAM-3, cell survival was decreased to 35% but the extent of survival induced by ifenprodil was not affected, precluding any deleterious effect of the UV light treatment on cell survival.Multiple comparisons were performed by two-way ANOVA with Bonferroni's correction; n.s., p > 0.05; *, p < 0.05; n = 5-8 batches of cultures per condition, in each culture 4-6 wells/condition.Data presented here (mean and SEM) were normalized to the control and NMDA conditions (see Methods for the calculation protocol).7A).For the orientation of aniline, in trajectories 01 and 02 we observed exchanges between 0 and 180°, whereas in trajectory 03 the angle stayed at 180°.On average, this angle is 60.8% at 180°, which means that this conformation is more stable by roughly 0.3 kcal/mol than the one at 0°.For the orientation of the azo moiety, in trajectory 01 the angle stayed at 180° and switched to 0° after 993 ns; for trajectory 02, we observed six conversions between the two basins; for trajectory 03, we observed some exchanges at the beginning, a stability from 125 to 810 ns, and then a final exchange.On average, this angle is 67.9% at 180°, which means that this conformation is more stable by roughly 0.4 kcal/mol than the one at 0°.At the end of this simulation without constraints, we observed that the distance between the GluN1 and GluN2B lower lobes had increased, far from the distance measured in the inhibited, full-length receptor (see Text S2).We therefore decided to perform new simulations where the protein heavy atoms were restrained close to their crystallographic positions (see Text S2 and Methods) (B) Evolution of the same angles but for the Rot-1 (grey) and Rot-2 (salmon) rotamers of OptoNAM-3 during the simulations under constraint.Left: orientation of the aniline.Right: orientation of the phenyl-azo moiety (C-C-N=N angle).(C) Evolution of the two angles for free trans-OptoNAM-3 in water.For the orientation of aniline, the two basins at 0° and 180° are populated respectively 48.2% and 51.8% of the time, whereas for the orientation of the phenyl-azo they are populated at

S34
Table S1.Summary of the IC 50 s of OptoNAMs in the dark and UV compared to the activity of their parent compounds.

Figure S1 .
Figure S1.Photochemical properties of OptoNAM-1 to -4 and their photodependent activity at GluN1/GluN2B receptors.(A-D) OptoNAM-1.(A) In solution, OptoNAM-1 can be switched from trans to cis configuration by UV illumination (365 nm) and back to trans by 440 nm light.(B) UV-visible absorption spectra of OptoNAM-1 in physiological aqueous buffer (Ringer pH 7.3, see Methods) in the dark (black curve), after 365 nm illumination (violet curve) and subsequent illumination of the 365 nm PSS by 440 or 490 nm light.Dashed line represents the wavelength of peak absorption of trans-OptoNAM-1 (320 nm).(C) OptoNAM-1 365 nm PSS (mostly cis) displays strong thermal stability in the dark in Ringer pH 7.3, at room temperature: no change of the absorption spectra was observed up to 60 min after 365 nm illumination.(D) Dose-response curves of OptoNAM-1 activity on GluN1/GluN2B receptors in the dark (black curve, IC 50 = 11 ± 1 µM, n = 4-6) or pre-illuminated with 365 nm (violet curve, IC 50 = 37 ± 2 µM, n = 5-17).(E-H) OptoNAM-2.(E) In solution, OptoNAM-2 can be switched from trans to cis by UV illumination (365 nm) and back to trans by 490 nm light.(F) UV-visible absorption spectra of OptoNAM-2 in physiological aqueous buffer (Ringer pH 7.3, see Methods) in the dark (black curve), after 365 nm illumination (violet curve) and subsequent illumination of the 365 nm PSS by 440 or 490 nm light.Dashed line represents the wavelength of peak

Figure S2 .
Figure S2.Decreased pKa of OptoNAM-1 and -2 compared to their parent compounds are likely responsible for their decreased activity.(A-F) pKa estimation of the trans (A,C,E) and cis (365 nm PSS) (B,D,F) isomers of OptoNAM-1 (A,B), OptoNAM-2 (C,D) and OptoNAM-3 (E,F) by UV-visible spectrum analysis.UV-visible absorption spectra were measured at different pH (1.7 in yellow, 5.5 in green, 7.3 in red and 11.4 in blue).Correspondence between spectra at different pH and OptoNAM protonated and unprotonated chemical structures is indicated.A leftward shift and a decrease of the absorbance peak corresponding to the aminopyridine carrying the charge was observed upon OptoNAM deprotonation.This effect is most obvious for OptoNAM-2 and characteristic of the deprotonated spectra of 2,6-diaminopyridine bases.9Based on this analysis, at physiological pH, OptoNAM-1 and -2, either in cis or trans, are unprotonated, while OptoNAM-3 is protonated.

Figure S3 .
Figure S3.Additional data relative to Figure 3. (A) Inhibition by OptoNAM-3 365 nm PSS is exclusively mediated by the remaining trans isomer still present in solution.(A) Dose-response curves of OptoNAM-3 activity on GluN1/GluN2B in the dark (black curve, IC 50 = 0.38 ± 0.03 µM, n = 4-21), pre-illuminated by 365 nm (purple curve, IC 50 = 1.7 ± 0.2 µM, n = 4-17) and theoretical dose-response curve (blue) of a mixture of 18% trans-and 82% cis-OptoNAM-3 (corresponding to the 365 nm PSS determined by HPLC) assuming that only the trans isomer is active.The theoretical curve was calculated from the trans dose-response curve in the dark.The 18% trans theoretical dose-response curve superimposes well to the one of OptoNAM-3 365 nm PSS.(B) Dose-response curves of OptoNAM-3 activity on GluN1/GluN2B in the dark (black curve, IC 50 = 0.38 ± 0.03 µM, n = 4-21), preilluminated by 350 nm (lavender curve, IC 50 = 4.4 ± 0.6 µM n = 3-13) and theoretical dose-response curve (pink) of a mixture of 9% trans-and 91% cis-OptoNAM-3 (corresponding to the 350 nm PSS determined by HPLC) assuming only the trans isomer is active.The 9% trans theoretical dose response curve is also well superposed to the one of OptoNAM-3 350 nm PSS.This shows that the activity of the 365 nm PSS and 350 nm PSS of OptoNAM-3 can entirely be explained by the amount of remaining trans isomer in the PSS.

Figure S4 .
Figure S4.OptoNAM-3 decreases NMDA-induced neuronal death in a photodependent manner.Percentage of neuronal survival in cultured cortical neurons exposed either to control (0.01% DMSO and 10 µM Glycine), NMDA (100 µM NMDA + 10 µM Glycine), NMDA + ifenprodil or NMDA + OptoNAM-3 (100 µM NMDA + 10 μM Glycine + 5 µM of inhibitor), in the dark (grey bar) or after 2 min UV (365 nm) illumination (violet bar).In the presence of trans-OptoNAM-3 (dark) or ifenprodil, cell survival increased to 50% and 60%, respectively.When 365 nm illumination followed the addition of OptoNAM-3, cell survival was decreased to 35% but the extent of survival induced by ifenprodil was not affected, precluding any deleterious effect of the UV light treatment on cell survival.Multiple comparisons were performed by two-way ANOVA with Bonferroni's correction; n.s., p > 0.05; *, p < 0.05; n = 5-8 batches of cultures per condition, in each culture 4-6 wells/condition.Data presented here (mean and SEM) were normalized to the control and NMDA conditions (see Methods for the calculation protocol).

Figure S5 .
Figure S5.OptoNAM-3 photomodulates Xenopus tadpole locomotion in vivo: protocol and tadpole locomotion normalized to baseline locomotion.Additional data relative to Figure 6.(A) Experimental design of the behavioral tests performed on tadpoles incubated in control (0.1% DMSO) or OptoNAM-3 at 5 µM and exposed to UV/green (or blue/green) light cycles.(B,C) Normalized distance (compared to baseline) traveled by tadpoles (1 point represents the mean distance traveled by the 3 tadpoles of one well) incubated in control (white bars and black points) or in 5 µM OptoNAM-3 (grey bars and points), in the dark and during UV/green light cycles (or blue-green light cycles for panel C).Note that UV light influences tadpole locomotion on its own.n = 16 wells for (B) and n = 17 wells for (C), which corresponds to a total of 48 and 51 tadpoles, respectively.The means of 3 tadpoles per well were used to conduct a paired statistical test.n.s., p > 0.05; ***, p < 0.001; **, p < 0.01; Friedman test followed by Dunn's multiple comparison test.

Figure S6 .
Figure S6.OptoNAM-3 photochemical properties in different solvents.(A) UV-visible absorption spectra of OptoNAM-3 (25 µM) in physiological aqueous solution (Ringer at pH 7.3), DMSO and toluene.Black curves represent the dark PSS, the violet ones OptoNAM-3 365 nm PSS, the blue and green curves represent OptoNAM-3 460 nm and 550 nm PSS obtained after illumination of the dark PSS, and the dotted blue and green curves represent OptoNAM-3 460 nm and 550 nm PSS obtained after illumination of the 365 nm PSS.(B) Maximum absorption wavelengths of the peaks corresponding to the π → π* transition (in regular police) and n → π* transition (in italic) for the different OptoNAM-3 PSS in the different solvents.For each solvent, the absorbance of each peak relative to the absorbance of the π → π* transition peak of the dark PSS is indicated in parenthesis.

Figure S7 .
Figure S7.Evolution of trans-OptoNAM-3 conformation in its binding site and in water.(A) Evolution of the two angles that describe bound trans-OptoNAM-3 conformations during 3 trajectories of a MD simulation without constraints.The three colors represent the three trajectories.Left: orientation of the aniline.Right: orientation of the phenyl-azo moiety (C-C-N=N angle) (Figure7A).For the orientation of aniline, in trajectories 01 and 02 we observed exchanges between 0 and 180°, whereas in trajectory 03 the angle stayed at 180°.On average, this angle is 60.8% at 180°, which means that this conformation is more stable by roughly 0.3 kcal/mol than the one at 0°.For the orientation of the azo moiety, in trajectory 01 the angle stayed at 180° and switched to 0° after 993 ns; for trajectory 02, we observed six conversions between the two basins; for trajectory 03, we observed some exchanges at the beginning, a stability from 125 to 810 ns, and then a final exchange.On average, this angle is 67.9% at 180°, which means that this conformation is more stable by roughly 0.4 kcal/mol than the one at 0°.At the end of this simulation without constraints, we observed that the distance between the GluN1 and GluN2B lower lobes had increased, far from the distance measured in the inhibited, full-length receptor (see Text S2).We therefore decided to perform new simulations where the protein heavy atoms were restrained close to their crystallographic positions (see Text S2 and Methods) (B) Evolution of the same angles but for the Rot-1 (grey) and Rot-2 (salmon) rotamers of OptoNAM-3 during the simulations under constraint.Left: orientation of the aniline.Right: orientation of the phenyl-azo moiety (C-C-N=N angle).(C) Evolution of the two angles for free trans-OptoNAM-3 in water.For the orientation of aniline, the two basins at 0° and 180° are populated respectively 48.2% and 51.8% of the time, whereas for the orientation of the phenyl-azo they are populated at

Figure S8 .
Figure S8.Additional data relative to Figure 7. (A) Superposition of the experimental π → π* and n → π* bands of the UV-Vis spectra of free OptoNAM-3 dark PSS (trans state, dark line) and 365 nm PSS (violet line) in aqueous buffer (Ringer pH 7.3), to the theoretical π → π* and n → π* transitions of OptoNAM-3 in implicit water computed by DFT calculations (blue bars representing the range of computed wavelengths for trans-OptoNAM-3 across the different snapshots of the dynamic, see Table S2).(B) Relationship between the C-C-N=N torsion angle (as highlighted in pale red in the inset chemical structure) of bound trans-OptoNAM-3 in the 11 snapshots selected for DFT calculations and their computed n→π* absorption wavelengths, for Rot-1 (in grey) and Rot-2 (in salmon).(C-E) Relationships between the N=N-C-C (C), C-N=N-C (D), and the C-C-N=N (E) torsion angles (as highlighted in pale red in the inset chemical structures) of bound trans-OptoNAM-3 in the 11 snapshots selected for DFT calculations and their predicted oscillator strengths, for Rot-1 (in grey), and Rot-2 (in salmon).Linear regressions: R 2 = 0.746 (C) and 0.706 (D).
Figure S8.Additional data relative to Figure 7. (A) Superposition of the experimental π → π* and n → π* bands of the UV-Vis spectra of free OptoNAM-3 dark PSS (trans state, dark line) and 365 nm PSS (violet line) in aqueous buffer (Ringer pH 7.3), to the theoretical π → π* and n → π* transitions of OptoNAM-3 in implicit water computed by DFT calculations (blue bars representing the range of computed wavelengths for trans-OptoNAM-3 across the different snapshots of the dynamic, see Table S2).(B) Relationship between the C-C-N=N torsion angle (as highlighted in pale red in the inset chemical structure) of bound trans-OptoNAM-3 in the 11 snapshots selected for DFT calculations and their computed n→π* absorption wavelengths, for Rot-1 (in grey) and Rot-2 (in salmon).(C-E) Relationships between the N=N-C-C (C), C-N=N-C (D), and the C-C-N=N (E) torsion angles (as highlighted in pale red in the inset chemical structures) of bound trans-OptoNAM-3 in the 11 snapshots selected for DFT calculations and their predicted oscillator strengths, for Rot-1 (in grey), and Rot-2 (in salmon).Linear regressions: R 2 = 0.746 (C) and 0.706 (D).

Table S2 .
Computed vertical energies and oscillator strengths of 11 snapshots of free trans-OptoNAM-3 in implicit water for the 2 first visible transitions using the B2PLYP functional.

Table S3 :
Computed vertical energies and oscillator strengths of 11 snapshots of bound trans-OptoNAM-3 inside the protein for the 2 rotamers and for the 2 first visible transitions using the B2PLYP functional.

Table S4 :
Computed vertical energies and oscillator strengths of snapshot 0 of bound trans-OptoNAM-3 for the 2 rotamers and for the 1st transition (n → π*), inside the protein (1, first line); without the protein and compound geometry frozen (2, second line), and without the protein after being optimized in vacuum (3, third line).